Legislative term limits and electoral disconnection

Questions or concerns about whether holding office would entice public officials into abusing their authority have afflicted many generations of thinkers. The earliest discussions on this subject date back to the Old Testament and ancient Greece. One solution to the power dilemma was identified in the imposition of legal restrictions preventing rulers from staying in office indefinitely (Petracca Reference Petracca1993). The debate around term limits was revived when Mexico’s Partido Nacional Revolucionario amended the 1917 Constitution to introduce a ceiling for the election of legislators in the 1930s. Other countries subsequently adopted this institution, including Costa Rica (1949), Malaysia (1957), Ecuador (1979), and the Philippines (1987). The term limit debate has also been a distinctive feature of US politics for a long time. Grassroots movements championing the implementation of term limit reforms go back to the early 1980s. Since the initial referenda in Colorado, Oklahoma, and California in 1990, these movements have grown considerably. As of 2022, term limits for representatives are in force in 16 states (see Table A1 in the online Appendix).

After its diffusion in the Americas and Asia, the issue of restricting the possibility of re-election for MPs made its way to Africa and Europe. In some cases, these restrictions have been proposed for national assemblies (e.g., France and Libya), while in other cases, they are currently in place at the national (e.g., Algeria and Rwanda) or sub-national level (e.g., Switzerland). The table below (Table 1) presents an overview of the characteristics of term limit rules across the world (including proposed reforms as of 2022). We classify the restrictions based on (a) whether they apply to the national or sub-national level (or both), (b) the extent of the limit (e.g., two consecutive terms), and (c) whether they apply to one or both chambers. We also trace their legal sources. As for the level, we introduce a third category, “party-based” (or “party-imposed”) limits. These are constraints enforced by individual parties rather than existing as a codified constitutional rule. This category captures our case study, the Five Star Movement in Italy, and some aspects of the Swiss system.Footnote ¹ As far as we are aware, the Five Star Movement is the only case of term limits for members of the national parliament currently in force in both chambers within the European continent.Footnote ²

Table 1. Term limits in parliament

*The 2018 Constitution draft establishes a limit of three terms for both chambers (Art. 51). The 2022 Constitution draft establishes a limit of two consecutive terms for both chambers (Art. 262).

**As of this writing, 16 states have term limits for legislators (see Table A1 in the online Appendix). Term limits were introduced in Utah and Idaho in 1994 and subsequently repealed. In four states (Massachusetts; Oregon; Washington; Wyoming), term limits were introduced between 1992 and 1994 and subsequently deemed unconstitutional by the respective State Supreme Court.

The Five Star Movement first ran in the 2013 parliamentary election, obtaining 25.6% of votes. The party debuted as a “populist challenger” party in the Italian political scene with strong Euroskeptic sentiment and a crusade against the Euro (Giannetti et al. Reference Giannetti, Umansky and Sened2024). It ran again in 2018 (32.7% of votes), becoming the party with the highest number of seats in the legislature. From 2018 to 2022, the M5S saw a decline in support. Throughout this period, the party governed with the right-wing party Lega first (Conte I government, 2018–2019), then with the center-left party PD (Conte II government, 2019–2021), and finally joined a national-unity government led by former ECB President Mario Draghi (2021–2022). In 2019, during the second Conte administration, the M5S held more moderate stances, reneging, for instance, some of the Euroskeptic proposals made in the early stages of its life. This evolution indicates a gradual “party institutionalization” of the Five Star Movement within the political system (Musella and Vercesi Reference Musella and Vercesi2019: p. 225).

Throughout the years, the Five Star Movement has been promoting term limits as one of its flagship internal policies, with one of its co-founders, Beppe Grillo, as the primary advocate. They are explicitly mentioned in the latest version of the party’s “Code of Ethics” (February 2023).Footnote ³ Specifically, Art. 2 of the Code of Ethics states that no candidate can run under the banner of the Five Star Movement if they have “already served two terms” (p. 2) for any elective position as a public official. Therefore, the 2022 election in September (19^th electoral period) was the first time Five Star Movement affiliates who had already served two mandates were unable to seek re-election. As a consequence, several prominent politicians linked to the party, such as former Justice Minister Alfonso Bonafede, Parliamentary Affairs Minister Riccardo Fraccaro, Infrastructure and Transport Minister Danilo Toninelli, and President of the Chamber of Deputies Roberto Fico, did not run for office, having already served as parliamentarians in the 17^th and 18^th electoral periods.

Grillo’s publicly stated rationale behind the adoption of term limits is that politics should be conceived as “civil service” and that measures must be taken to prevent citizens from becoming “professional” politicians working on behalf of the lobbies.Footnote ⁴ MPs should be mere “employees” whose job is strictly to implement the party program and should get fired if they fail to do so (Casaleggio and Grillo Reference Casaleggio and Grillo2011). Broadly, this is typical of challenger parties that often claim to be “untarnished by office” (Hobolt and Tilley Reference Hobolt and Tilley2016: p. 974). Overall, term limit proposals are a global phenomenon that, both in Italy and elsewhere, seems to find its roots in citizens’ distrust of traditional party politics and in parties’ desire to turn parliamentarians into party delegates. The reasons put forward by the Five Star Movement in support of legislative term limits largely reflect those put forward by advocates in other political contexts. In countries like Algeria (Khelfa and Zamani Reference Khelfa and Zamani2023), Chile (Heiss Reference Heiss2021), Costa Rica (Carey Reference Carey1996), Peru (Alfaro Reference Alfaro2021), and the Philippines (Labonne, Parsa and Querubin Reference Labonne, Parsa and Querubin2021), these restrictions were in fact introduced as an attempt to enhance the credibility of political elites perceived as corrupt or prevent the emergence of political dynasties.

The diffusion of term limits prompted political scientists to dig into the relationship between electoral incentives and the functioning of legislatures (for a literature review, see Mooney Reference Mooney2009). Of the several consequences of adopting these restrictions identified by the existing literature, the behavioral effect is among the most salient ones. When members of the legislative body are constrained by the presence of term limit rules, building a long-running parliamentary career can become a much harder – or even impossible – goal to attain. Hence, elected officials have fewer incentives to specialize and perform constituency-specific or particularistic activities (Klesner Reference Klesner, Baturo and Elgie2019). Our focus is on electoral disconnection. Carey’s (Reference Carey1996) work called attention to the fact that the incentives to allocate resources to strategic parliamentary activities and comply with party discipline shrink when parliamentarians become aware that they cannot seek re-election in sub-sequent terms. In other words, legislators grow electorally disconnected, that is, more inclined to forgo electorally rewarding undertakings. For instance, Carey shows that, following the adoption of term limits, Costa Rican legislators became more independent when it came to complying with party directives in roll-call votes. More recently, Fouirnaies and Hall (Reference Fouirnaies and Hall2022) suggested that term-limited state representatives in the US tend to decrease their productivity, although this effect does not necessarily apply to female representatives (Holman and Mahoney Reference Holman and Mahoney2023).

Disconnection happens because term-limited legislators in their last term no longer need to please the “selectorate” on which their tenure would depend if re-election were a possibility. Constrained by term limit rules, individual MPs become less responsive to and detached from the demands and inputs of their selectorate, which translates into changes in parliamentary behavior and activities. Additionally, the mechanisms party leaders normally resort to in order to mitigate the agency problems (slacking and shirking) that inevitably arise in representative democracies (Strøm Reference Strøm2000) lose effectiveness because candidacy can no longer be used as an incentive. These mechanisms range from screening potential candidates to closely monitoring members’ performance in parliament (Müller Reference Müller2000; Cox and McCubbins Reference Cox and McCubbins2007).

The notion of electoral disconnection is coherent with rational choice theory surmising that parliamentarians’ behavior is primarily driven by the goal of re-election (Fenno Reference Fenno1973; Mayhew Reference Mayhew1974) and shaped by institutions and formal rules (Strøm Reference Strøm1997), such as the electoral system and candidate selection procedures. Specifically, when the electoral procedure is candidate-centered (e.g., first-past-the-post systems or open-list PR), representatives have more incentives to cultivate “personal vote” strategies (Cain, Ferejohn and Fiorina Reference Cain, Ferejohn and Fiorina1987; Carey and Shugart Reference Carey and Shugart1995). Instead, when the electoral procedure is party-centered, i.e., the selection of candidates is heavily or solely influenced by party leaders (e.g., closed-list PR), the greatest incentive is to please the leaders (e.g., Müller Reference Müller2000) and adhere to the party line in roll-call divisions and in performing legislative activities. Drawing upon this research, we likewise assume that MPs are “single-minded re-election seekers” (Mayhew Reference Mayhew1974) and that their behavior is driven by the intent of maximizing votes (and thus their chances of re-election) and that parliamentary activities are exploited “strategically” as means to pursue these objectives (Strøm Reference Strøm1997).

There exists, of course, the perspective that parliamentarians also pursue policy and office aspirations transcending re-election concerns (e.g., Searing Reference Searing1994). That said, in justifying our assumptions, we concur with Heitshusen, Young and Wood (Reference Heitshusen, Young and Wood2005). With respect to whether re-election should be taken as the main objective of all legislators, the authors suggest that we should consider the re-election goal as a constraint. Representatives may indeed run for office for reasons other than purely securing office (e.g., policy ambitions), but the necessity to stay in office to be able to go after any other goal they might have will impose itself as a critical and ineludible constraint. Only when confident about the odds of re-election can parliamentarians who are more policy- or constituency-oriented allocate all of their resources to these ends. However, since re-election in the next term is not 100% guaranteed for any member of the legislature, MPs will be forced to redirect at least some – if not most – of their resources toward re-election. Re-election, therefore, becomes the conditio sine qua non any other political aspiration can be pursued. This is especially true in the Italian case, as indicated by Chiaramonte et al. (Reference Chiaramonte, Emanuele, Maggini and Paparo2022). In the last decade, Italy has been characterized by high electoral volatility and instability in both political supply and demand, where parties that emerged as winners in one election often saw their electoral fortunes shrink significantly in the next round.Footnote ⁵

With this in mind, to quantify potential changes in electoral incentives, we look at if and how levels of election-related activities change as a result of the term limit restrictions. There is no set list of parliamentary activities that are always strategic for re-election because the strategic nature of these activities depends on the existing institutions and formal rules. In the next section, we contextualize the environment in which the Five Star Movement operates, and we derive testable hypotheses related to strategic parliamentary activities in Italy.

The Five Star Movement and strategic parliamentary activities in Italy

We investigate the electoral disconnection hypothesis by studying if and how the term limits imposed by the Five Star Movement in Italy affected the incentives of legislators to allocate resources to strategic election-related activities. As we anticipated, the strategic nature of parliamentary activities depends on the existing institutions and formal rules, such as the electoral system and candidate selection procedures.

In the Italian case, in the 2013 election (17^th electoral period), all seats were allocated based on a closed-list proportional (PR) system with a majority bonus (“Legge Calderoli,” “Calderoli” Electoral Law). For the 2018 and 2022 elections (18^th and 19^th electoral periods), Italy used a mixed system for both chambers (“Legge Rosato,” “Rosato” Electoral Law). In this case, the country was divided into single and multi-member districts (SMD/MMD), where 63% of seats were allocated based on proportional representation (closed-list for domestic constituencies), and the remaining 37% of seats were allocated in a majoritarian fashion. Majoritarian candidates would have to be endorsed by one or more political parties that formed a pre-electoral coalition. Voters would then jointly vote for an electoral list for PR seats and the candidates corresponding to that list for SMD seats, precluding the possibility of split-ticket voting.Footnote ⁶

With respect to candidate selection procedures, the Five Star Movement is somewhat anomalous within the Italian party scene. In 2013, the lists of candidates for the upcoming election were in part the product of online primaries. Only individuals with no official party affiliation and who had not held public office between 2009 and 2013 would be able to feature on these lists. Moreover, voters in the online primaries were only able to choose from candidates residing in their own electoral district (i.e., no “parachuted” candidates). Importantly, these candidates still had to be screened and pre-approved by the party before they could run in these primaries. In 2018, the same method was used for assembling the PR lists. As for majoritarian elections, the candidates were unilaterally selected by Five Star Movement leaders, allowing for the possibility of parachuting candidates in single-member districts.Footnote ⁷ The selection process was similar in 2022. Given the characteristics of the electoral systems discussed so far, in the periods covered in this study (2013–2022), candidacy for affiliates of the Five Star Movement was primarily determined by the party leaders but also by its electorate. To varying degrees, candidates aspiring to be elected had to secure the support of both.Footnote ⁸ To test the extent to which the electoral disconnection hypothesis holds, we consider a series of parliamentary activities that can be used for signaling purposes vis-à-vis both party leaders and constituents.

The first activity we consider, namely, our first dependent variable, is the number of “rebel” votes, which are relevant to both selectorates. Rebel votes are the instances where representatives vote against the party line in a given roll-call division. The propensity of individual legislators to defect largely depends on the form of government. In a parliamentary system like Italy, executive survival is conditional on party unity because a legislative majority must support the executive at all times. The opposition likewise needs to build a unified coalition to obstruct and make credible removal threats. As a result, defection is relatively rare in the context of parliamentary government (Sieberer Reference Sieberer2006). To ensure unity party leaders may employ various carrot-and-stick tools, such as floor time privileges or committee appointments (see Andeweg and Thomassen Reference Andeweg and Thomassen2011). Among these, candidacy is the most important one. Hence, ambitious party affiliates who wish to run for office will be discouraged from voting against the party line because of the negative signal “rebelling” sends to the leaders responsible for candidate selection.

The propensity of individual legislators to defect also depends on the degree to which the electoral system provides incentives for assembly members to pursue personal vote strategies. As discussed above, Italy is a mostly party-centered case in that the selection of candidates for national elections – both in the case of PR and SMD elections – is heavily influenced or entirely determined by party elites (e.g., D’Alimonte Reference D’Alimonte, Matthew and Martin2003). Furthermore, the literature has demonstrated that, in general, Italian parliamentarians who vote against the party directives more frequently are less likely to be re-elected (Galasso and Nannicini Reference Galasso and Nannicini2017).Footnote ⁹ In the specific context of the Five Star Movement, this relationship could be in part moderated by the presence of online primaries. Albeit marginally, these primaries allow the public to influence the selection of candidates. Because of the active involvement of voters in the process of candidate selection, primaries might create at least some incentives for legislators elected through PR lists to cultivate personal vote and disregard the official party position regarding a particular division in parliament. In other words, primaries might make it rational for some representatives to “break the whip” (Willumsen and Öhberg Reference Willumsen and Öhberg2017; Rehmert Reference Rehmert2020). However, scholars have shown that this was not the case for Five Star Movement affiliates during the 17^th electoral period (Rombi and Seddone Reference Rombi and Seddone2017). Moreover, the Five Star Movement was also the parliamentary group with the highest number of expulsions of its own members from the party. Oftentimes, these expulsions were the result of individual legislators criticizing party leaders, which might have reduced potential incentives to deviate springing from the direct involvement of the public in the selection of candidates (Tronconi Reference Tronconi2018; Vittori Reference Vittori2023). To summarize, even in the case of Five Star Movement candidates running in the online primaries, candidacy still had to undergo a screening process by the party, and candidacy for majoritarian elections was exclusively determined by the party. Finally, the act of rebelling might have been punished with measures as drastic as expulsion. It follows that rebel votes would have been an unwise move for Five Star Movement parliamentarians hoping to be re-elected. When MPs are constrained in their ability to be re-elected by the presence of term limits, though, the traditional mechanisms that foster party discipline lose effectiveness and the incentives for legislators to vote cohesively to please the party elites dwindle. Therefore, term-limited representatives in their second term might increase reliance on rebel votes to further personal policy (position-taking behavior) and career ambitions. Indeed, one possible explanation for this phenomenon, as suggested by Carey (Reference Carey1996), is that candidates who plan on pursuing a career outside of parliament could exploit legislative votes to send a message or appeal to certain interest groups or lobbies. It could also be that term limits induce a “Burkean shift,” as hypothesized by Carey et al. (Reference Carey, Niemi, Powell and Moncrief2006), that is, a shift from a party delegate to a more “trustee” style of representation.

Defining rebel votes as a legislator’s percentage ratio of votes that deviate from the official position of their party in a given roll-call division to the total votes cast by the legislator, we hypothesize that:

The second parliamentary activity we focus on (our second dependent variable) is the number of plenary speeches made by individual representatives. In party-centered electoral systems in particular, the possibility of speaking in plenary meetings is typically controlled by party leaders (Proksch and Slapin Reference Proksch and Slapin2012) and constrained by institutional rules (Giannetti and Pedrazzani Reference Giannetti and Pedrazzani2016). The number of speeches made by an MP is affected by different factors, such as the ideological distance between a legislator and the party elites, the level of expertise of parliamentarians with respect to a specific topic, or the level of experience as a member of a relevant parliamentary committee (Yildirim et al. Reference Yildirim, Kocapınar and Ecevit2019; Louwerse and Van Vonno Reference Louwerse and Van Vonno2022). Speeches are an important channel for candidates to reaffirm their allegiance to the party (Bauman et al. Reference Baumann, Debus and Klingelhöfer2017). They are also especially important for candidates running in primary elections because parliamentary speeches offer a good opportunity for politicians to send a message to voters, given that these speeches are generally picked up by the media and circulated on social networks. In fact, Marcinkiewicz and Stegmaier (Reference Marcinkiewicz and Stegmaier2019) show a positive relationship between the number of speeches in parliament and the number of preferential votes obtained by a candidate. In the Italian case, the control exercised by parties on the possibility of speaking in a plenary meeting is dictated by the specific procedure in use (e.g., decree-laws), and it is not exceptionally tight. Proksch and Slapin (Reference Proksch and Slapin2015) find Italy to be a median case in the “party leadership control over floor access” (p. 83), with parliamentarians generally free to speak in plenary sessions. In fact, Italy is classified as a “party list favored, individual access” type, halfway between the two extremes “individual access, no party list” and “party list, no individual access.”Footnote ¹⁰ Considering the benefits legislators can derive from parliamentary speeches highlighted by the literature reviewed so far, for Five Star Movement representatives (both those going through primaries and those selected by party leaders for majoritarian races), speeches might be considered a strategic activity for re-election. Consequently, legislators who cannot seek re-election due to term limits will have fewer incentives to speak. Term-limited members might choose to disengage because the ban on re-election results in lower utility that can be extracted from performing this activity.

Defining parliamentary speeches as the number of plenary speeches made by an individual assembly member, we hypothesize that:

In addition to rebel votes and plenary speeches, we look at two variables more strictly pertaining to legislative activity and less procedurally constrained (compared to speeches): legislative proposals (private members’ bills) and written parliamentary questions, which Italian parliamentarians can submit freely. Both of these activities have been used by scholars to measure the productivity of MPs (Akirav Reference Akirav2019). Introducing many bills might boost the odds of being selected when seeking re-election because presenting proposals can be perceived as partisan “service” by the party leaders in charge of selecting candidates (Hermansen Reference Hermansen2018; Put et al. Reference Put, Benjamin De Vet, Wauters and Maddens2024). Furthermore, looking at the case of New Zealand, Williams and Indridason (Reference Williams and Indridason2018) explicitly argue that legislative proposals introduced by members of the legislature serve as an electoral connection.

In the Italian context, parliamentary questions are instead primarily used to promote territorial interests (Russo Reference Russo2013), as well as a tool to scrutinize and hold the government accountable. Written questions can help representatives as a means for them to “make themselves known” to their constituents and potentially gain votes in primary elections. This is because parliamentary questions can be used to bring local issues to the attention of the national government and demand its intervention. Moreover, due to the succinct nature of these questions, they are easy to publicize in the local media such that voters can be made aware that their MP is “fighting” to solve problems that affect them on a daily basis (Russo Reference Russo, Martin and Rozenberg2012).

While we remain a priori agnostic when it comes to testing the electoral disconnection hypothesis by looking at both bills and written questions, we note earlier studies have found no evidence that proposals can help candidates be re-elected in Italy (see also Table A2 in the online Appendix). Conversely, parliamentary questions have been found to be a strategic tool for re-election, but mostly for incumbents in the Southern regions, which have a stronger tradition of preferential voting (Marangoni and Russo Reference Marangoni and Russo2018). Following the same logic as plenary speeches, bills and questions can be beneficial for members trying to secure the support of the public and party leaders, but they are costly undertakings. Parliamentarians constrained by term limits will not be able to extract the same utility as their non-term-limited colleagues from performing these activities.

Defining legislative proposals and parliamentary questions as the number of, respectively, private members’ bills and written parliamentary questions submitted by an individual MP, we hypothesize that:

Figure 1 summarizes our hypotheses and expectations.

Figure 1. Term limits and electoral disconnection.

The identified variables (rebel votes; speeches; proposals; questions) are deemed crucial when it comes to candidate selection and re-election by research cutting across different political systems. While our discussion mainly revolved around the electorally lucrative uses of these parliamentary activities, we acknowledge that these activities can serve multiple purposes. For example, in addition to electoral incentives, plenary speeches and bills can be exploited to signal policy priorities or to show off one’s capabilities with respect to holding a legislative office. Strictly speaking, we cannot necessarily directly observe or measure the underlying incentives (certainly not in the case of aggregate data). However, under the assumption that rational legislators will allocate resources to those activities that maximize the chances of re-election before anything else and that the activities we consider are paramount to this end, we can look at the extent to which the presence of term limits and thus electoral disconnection causes shifts in the way these activities are performed. In other words, in an aggregate sense, we think of speeches, proposals, questions, and (rebel) votes as “strategies” or, as put by Strøm (Reference Strøm, Blomgren and Rozenberg2012), “endogenous prescriptions as to how actors (here: parliamentarians) may most successfully and efficiently act to maximize the likelihood of whatever outcomes they favor” (p. 87).

Given that our focus is strictly on the Italian case, we corroborate the validity of our variable selection empirically by estimating a logistic model of the re-election of candidates in the 17^th-18^th and 18^th-19^th electoral periods. Similar to previous studies (Marino and Martocchia Diodati Reference Marino and Martocchia Diodati2017; Marangoni and Russo Reference Marangoni and Russo2018), the results confirm that all the aforementioned parliamentary activities exhibit a significant impact on the probability of re-election, except for legislative proposals (see Table A2 in the online Appendix). Specifically, parliamentarians who made a higher number of plenary speeches or submitted more written questions were more likely to be re-elected in the next term. Conversely, those with higher rates of rebel votes were less likely to be re-elected. In the next section, we discuss our data and research design.

Data and research design

To test the electoral disconnection hypothesis, we investigate whether representatives affiliated with the Five Star Movement advancing from their first to their second mandate exhibit any differences in the levels of electorally lucrative parliamentary activities (speeches, bills, and written questions) and rebel votes, compared to other members in the legislature whose ability to run for office is not restricted by party-imposed term limits. Given the impossibility of running for election, we expect term-limited legislators serving their second mandate to be more likely to disregard parliamentary activities that are strategically performed by parliamentarians to increase the odds of securing candidacy and re-election and be more likely to rebel as the incentives to vote along the party lines fade. As we mentioned earlier, the Five Star Movement first gained seats in the Italian legislature in the 2013 election. Therefore, our study incorporates the last two electoral terms, the 17^th (2013–2018) and 18^th (2018–2022) period. We adopt a quasi-difference-in-differences approach to conduct an empirical analysis of the relationship between term limits and electoral disconnection. The basic model we estimate is the following:

$${y_{it}} = \alpha + {{\bf{x}}_{it}}\beta $$

where y _it is the dependent variable, namely, measures of parliamentary activities (speeches, questions, etc.), α is a constant, and x_it is a vector of predictors with coefficients β. We allow the intercepts to vary by legislator and region to capture unit-specific differences in the outcome. We include an interaction between a dummy variable identifying the MPs affiliated with the Five Star Movement and a dummy variable identifying the 18^th electoral period. The coefficient of this interaction term is our main effect of interest, denoting the change in behavior of Five Star Movement deputies between the two electoral periods covered in our study compared to the change in behavior of their colleagues who are not constrained by term limits.

We include the following controls: candidate vulnerability, electoral system (PR or SMD), legislative office, government membership, party switch, election chamber (House or Senate), experience in parliament (tenure), age, and gender. The model features an interaction term between candidate vulnerability and the electoral system variable to differentiate the vulnerability of legislators elected in majoritarian elections from that of legislators elected through PR lists. In what follows, we describe these variables in greater detail.

Candidate vulnerability, electoral system: Winning a parliamentary seat by a narrow margin in a majoritarian system or a proportional representation system makes an incumbent concerned about re-election chances. To increase their odds of securing a seat, MPs could try to increase their productivity (Akirav Reference Akirav2020), especially those furthering particularistic interests in the case of SMD candidates. The electoral system for the 2013 general election was a closed-list PR. In 2018, Italy switched to a mixed system (SMD and PR). To calculate electoral vulnerability, we followed Ferrara (Reference Ferrara2004)’s analysis of the “Mattarella” electoral law in Italy (the mixed electoral system in force from 1993 to 2005). For candidates elected based on proportional representation, we considered the number of candidates elected from a party list in each constituency and the position of the elected candidates on the list, i.e., whether an MP was first, second, third, etc., on a given list. For those elected in single-member districts, we calculated the gap between the top two candidates. The resulting indicator, thus, measures electoral volatility as the ratio between a candidate’s position on the list and the total number of candidates elected from that list for PR legislators and the gap between two candidates for SMD legislators.Footnote ¹¹ We then transformed the variable such that vulnerability for SMD and PR candidates would be on the same scale and that a one-unit increment would represent an increase in vulnerability in both cases.

Government membership, legislative office, party switch, election chamber: The intensity and quality of parliamentary activities for cabinet members and parliamentarians are fundamentally different, even among co-partisans. For instance, cabinet members do not submit written questions as questions are directed toward them. As such, we control for government membership. In the same vein, as demonstrated by Sorace (Reference Sorace2018), when activities are procedurally constrained or highly visible, such as plenary speeches, MPs holding a legislative office (whips, committee chairs, etc.) are more likely to engage in these activities due to expertise, status, or other political considerations. Therefore, we include a dummy for legislative office.Footnote ¹² We also add a control for representatives who switched parties during the electoral term, as their incentives might differ from those who did not switch. Some Five Star Movement MPs serving their second mandate might not have altered the intensity of parliamentary activities in the prospect of changing party affiliation to stay in office. Finally, we include a dummy recording the parliamentary chamber in which politicians served their mandate to account for potential variation in behavior that might be due to differences between the two chambers (e.g., Pinto Reference Pinto2023).

Tenure, age, gender: Experience in parliament (tenure), age at the time of election (age), and gender were also included as control variables. Higher parliamentary experience can make MPs more independent and capable of carrying out parliamentary tasks more efficiently (Kousser Reference Kousser2005). Additionally, higher tenure normally implies higher chances of being re-elected (Marino and Martocchia Diodati Reference Marino and Martocchia Diodati2017). Women may also be more productive in the legislature to “meet the expectations” behind their election (Holman and Mahoney Reference Holman and Mahoney2023), even if they may be barred from speaking in plenary meetings by their male counterparts (Fernandes, Lopes da Fonseca and Won Reference Fernandes, Lopes da Fonseca and Won2021). At the same time, the older a deputy is, the more likely they will not seek re-election. For each representative, we collected data related to the aforementioned socio-demographic characteristics (tenure, age, gender), election features (constituency, electoral system, position on the list, percentage of votes obtained by the candidate), and legislative behavior (number of plenary speeches, number of bills introduced, number of written questions submitted, proportion of rebel votes). We did this for both periods and both chambers of the Italian Parliament. In total, the dataset covers over 2000 assembly members. The population we consider for our study is those elected to either chamber in 2013 who started a second mandate in 2018.Footnote ¹³ This allows us to compare how each legislator’s behavior in parliament changed over time and whether this change is statistically different for MPs with term limit constraints and those without. The final product is a longitudinal parliamentarian-electoral period structure. In total, we have 317 legislators monitored over two terms (N = 634). Throughout this time, six governments in total held office: Letta (2013–2014), Renzi (2014–2016), and Gentiloni (2016–2018) in the 17^th period, Conte I (2018–2019), Conte II (2019–2021), and Draghi (2021–2022) in the 18^th period.

We obtained the data on the socio-demographic characteristics of elected officials and their legislative behavior via a SPARQL endpoint query of the official Linked Open Data repositories of the Italian Chamber of Deputies and Senate.Footnote ¹⁴ We collected the election data from Eligendo, the official data repository of the Italian Interior Ministry.Footnote ¹⁵ Finally, data on rebel votes were kindly supplied by the Openpolis foundation.Footnote ¹⁶ Table 2 presents some summary statistics of the variables included in our model. In the next section, we present our results and discuss their implications.

Table 2. Summary statistics

Additional tests and alternative explanations

In this section, we run some additional tests and present additional evidence to substantiate the mechanisms we propose and discount some potential alternative explanations. In the previous section, we presented the product of the estimation of a series of linear mixed-effect models with the inclusion of random intercepts for legislators and regions. To corroborate the robustness of our estimates, we also estimated negative binomial (count) models for proposals, speeches, and questions and a quasi-binomial (fractional logit) model for rebel votes as an alternative. These results are available in the online Appendix (Table B1 and Table B2) and are consistent with those shown in Table 3.

As an additional robustness check, we replicated our models by restricting the sample to each chamber individually. The results are available in Table B3 in the online Appendix and show no substantial differences when considering only deputies (lower house) or senators (upper house). The only exception we detect is that our main effect of interest is no longer significant in the case of rebel votes in the Senate.

We performed another robustness check. One possibility is that the changes we observe might be driven by the majority vs. opposition status, which the aggregate nature of the data in our main models does not allow us to deal with dynamically. For legislative proposals, parliamentary speeches, and parliamentary questions, we were able to disaggregate the data by cabinet.Footnote ²⁰ We thus compiled a new dataset where, for each member of our population, we recorded the levels of a given parliamentary activity by government (six governments in total). This allows us to explicitly take into account the majority and opposition status and party affiliation of each parliamentarian for each governmental period. We replicated the models presented in Table 3 with the inclusion of a dummy for majority status and cabinet random effects. The estimates are available in Table B4 in the online Appendix and are consistent with our main results.

As discussed above, one concern is that the significant changes in behavior we observe is the fact that the Five Star Movement transitioned from being an opposition party in the 17^th electoral period to joining the governing coalition in the 18^th period as well as the fact that the results might be driven by differences between the Five Star Movement and the other parties. To further rule out these possibilities, we compare the levels of parliamentary activity of Five Star Movement MPs serving their second mandate and those serving their first mandate in the 18^th electoral period, namely, after the party joined the government coalition. If it is true that changes in behavior were driven by the transition from opposition to majority status and electoral disconnection did not affect the incentives of term-limited legislators, we should not observe any meaningful differences between the group of Five Star Movement affiliates who can run for a second mandate and the group of those who cannot in the 18^th period. However, Table 4 shows the opposite. Two sample t-tests suggest that Five Star Movement MPs serving their second mandate submit fewer speeches and written questions and are more prone to rebelling, on average, than Five Star Movement MPs serving their first mandate in the same period. These differences are statistically significant at the conventional levels, except for rebel votes. The figures related to legislative proposals are consistent with the fact that we do not find private members’ bills to significantly impact the probability of re-election. These tests also speak to the heterogeneity between parties: in this case, we are comparing M5S legislators to other M5S legislators, meaning that the statistically significant differences that we detect cannot be driven by fundamental differences between the Five Star Movement and the other parties in the Italian Parliament.

Table 4. Average differences between Five Star Movement legislators serving first vs. second mandate in the 18^th electoral period

Note: Welch Two Sample t-tests; ***p < 0.01, **p < 0.05, *p < 0.10.

Going back to the incentives that drive parliamentary behavior, the electoral disconnection hypothesis and our analysis are built on the assumption that legislators are single-minded seekers of re-election, implying that other political goals become secondary to the re-election constraint. The increase in rebel votes that we observe is consistent with the fact that, no longer limited by re-election constraints, term-limited parliamentarians have fewer incentives to follow party directives as opposed to personal (or constituency) policy preferences in the event of misalignment between these preferences and the party line. Term limits could induce a “Burkean shift,” as hypothesized by Carey et al. (Reference Carey, Niemi, Powell and Moncrief2006), that is, a shift from a party delegate to a more “trustee” style of representation. It follows that non-electoral goals become drivers of MPs’ voting behavior once the electoral incentives are lost due to term limit restrictions. An alternative is that legislators were primarily interested in policy rather than re-election all along. If that were the case, we would expect to see the observed increase in rebel votes to be correlated with higher policy activism. To investigate this possibility, we run a simple test. We collected data on Italian representatives serving as “rapporteurs,” which implies high levels of interest and engagement in policymaking in parliamentary committees. We looked at the correlation between rebel votes and serving as a rapporteur in term-limited M5S MPs in the 18^th electoral term and in their non-term-limited colleagues.Footnote ²¹ We find a positive, albeit weak, correlation between the two in both cases, but the coefficients are not statistically different, meaning that the Five Star Movement does not exhibit an unusual behavior that might be explained by term limit restrictions in this regard (see Table B5 in the online Appendix). Thus, rebel votes do not appear to be linked to an increased interest in policy, which would not be coherent with the idea of disengagement that comes with electoral disconnection. Although we agree with scholars such as Strøm (Reference Strøm1997) that reducing parliamentary behavior to the single-minded re-election goal is a “simplification of reality” (p. 172), this result makes us more confident that re-election is, at the very least, at the top of representatives’ concerns rather than policy.

Another possibility is that changes in behavior might be due to at least some Five Star Movement legislators anticipating that they would switch parties. By changing party affiliation, Five Star Movement MPs would no longer be constrained by term limit rules. Consequently, they might carry out strategic parliamentary activities regularly to further re-election goals rather than losing the incentives to do so as posited by the electoral disconnection hypothesis. To rule out this possibility, we compared the levels of parliamentary activity of Five Star Movement affiliates serving their second mandate (18^th electoral period) who switched parties to those who did not. If party switching is driving the observed changes in behavior, the levels of our outcomes of interest should be significantly different across the two groups. However, Table 5 (two sample t-tests) suggests that, while representatives who switched tend to exhibit slightly higher levels of electorally lucrative activities and rebel votes, the differences between the two groups are not statistically significant at the conventional levels. In general, it is plausible to think that our main results suggesting that term-limited members exhibit higher rates of rebel votes might be linked to party switching. Specifically, MPs interested in joining a different party once they can no longer run in affiliation with the Five Star Movement might start rebelling in roll-call votes to position themselves closer to the party they intend to join in the future (Nokken Reference Nokken2000; Meneghetti et al. Reference Meneghetti, Pacini, Dal Monte, Cracchiolo, Rossi, Mazzoni and Micera2023). Yet, we find that the average level of rebel votes is almost identical across the two groups, making us more confident that the observed change in behavior is not affected by party switching.

Table 5. Average differences between Five Star Movement party switchers and non-switchers in the 18^th electoral period

Note: Welch Two Sample t-tests.

Overall, these results are coherent with our theoretical framework for which we expect term-limited legislators to become electorally disconnected when they can no longer seek re-election.